Granted Patent B2

US 12,493,461 · App. 18/273,282 · Granted Dec 9, 2025

Inventors: Hasan Sozer (Istanbul, TR); Ali Ekici (Istanbul, TR); Okan Orsan Ozener (Istanbul, TR)

Assignee: OZYEGIN UNIVERSITESI

G06F8/71G06F8/433

Abstract

Claims (42)

1 . A method for providing recovery of modular software architecture by means of a processor unit configured to access a source code repository comprising source codes of software consisting of a plurality of modules and configured to access a storage database comprising an update history of source codes of the software, the method comprises the process steps of:

accessing the source code repository,

detecting predetermined references in the codes of different modules in the source code repository,

detecting different modules containing same references,

determining a structural dependence between modules based on a number of same references contained in the different modules containing same references,

accessing the storage database,

detecting integrated updates whose update times are in predetermined proximity,

detecting integrated modules included together in the integrated updates,

determining an evolutionary dependency between modules based on a number of integrated updates in which the integrated modules are present together,

based on the structural dependence between the modules, clustering the modules in a predetermined number of clusters, with each module is in one cluster and forming a structural dependency cluster,

based on the evolutionary dependence between the modules, clustering the modules in a predetermined number of clusters, with each module is in one cluster and forming an evolutionary dependency cluster,

combining the structural dependency cluster and the evolutionary dependency cluster based on predetermined weight coefficients and obtaining a final cluster, and

performing, by the processor unit, modular software architecture recovery based on the final cluster.

2 . The method according to claim 1 , wherein the references are method/function calls or variable accesses.

3 . The method according to claim 1 , further comprises the process steps of:

detecting textual similarities of the codes of the different modules in the source code repository,

determining a semantic dependency between modules based on a textual similarity ratio,

based on the semantic dependence between the modules, clustering the modules in a predetermined number of clusters, with each module is in one cluster and forming a semantic dependency cluster, and

combining the structural dependency cluster, the evolutionary dependency cluster and the semantic dependency cluster based on predetermined weight coefficients and obtaining the final cluster.

4 . The method according to claim 1 , further comprises the process steps of:

detecting accesses between the different modules in the source code repository during at runtime,

determining a dynamic dependency between modules based on the accesses,

based on the dynamic dependence between modules, clustering the modules in a predetermined number of clusters, with each module is in one cluster and forming a dynamic dependency cluster, and

combining the structural dependency cluster, the evolutionary dependency cluster and the dynamic dependency cluster based on predetermined weight coefficients and obtaining the final cluster.

5 . The method according to claim 1 , further comprises the process steps of:

detecting access of the different modules in the source code repository to common components,

determining an external dependency between modules based on common accesses,

based on the external dependence between the modules, clustering the modules in a predetermined number of clusters, with each module is in one cluster and forming an external dependency cluster, and

combining the structural dependency cluster, the evolutionary dependency cluster and the external dependency cluster based on predetermined weight coefficients and obtaining the final cluster.

6 . The method according to claim 5 , wherein the common components are a variable, a function, a database table or a resource.

7 . The method according to claim 1 , further comprises the process steps of:

detecting textual similarities of the codes of the different modules in the source code repository,

determining a semantic dependency between modules based on a textual similarity ratio,

based on the semantic dependence between the modules, clustering the modules in a predetermined number of clusters so that each module is in one cluster and forming a semantic dependency cluster,

detecting accesses between the different modules in the source code repository at runtime, determining a dynamic dependency between modules based on the accesses,

based on the dynamic dependence between the modules, clustering the modules in a predetermined number of clusters so that each module is in one cluster and forming a dynamic dependency cluster,

detecting access of the different modules in the source code repository to common components,

determining an external dependency between modules based on the common accesses,

based on the external dependence between the modules, clustering the modules in a predetermined number of clusters, with each module is in one cluster and forming an external dependency cluster, and

combining the structural dependency cluster, the evolutionary dependency cluster, the semantic dependency cluster, the dynamic dependency cluster and the external dependency cluster based on predetermined weight coefficients and obtaining the final cluster.

8 . The method according to claim 1 , wherein the source code repository and the storage database are provided in a memory unit.

9 . The method according to claim 1 , wherein the processor unit accesses the source code repository and the storage database via a data bus.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (20)